Differentiable Programming of Chemical Reaction Networks

TL;DR

This paper introduces a differentiable framework for chemical reaction networks (CRNs) that enables training to discover sparse networks capable of performing complex computational tasks, including oscillators and chemical computing devices.

Contribution

It presents a novel differentiable formulation of CRNs that allows for optimization and discovery of functional reaction networks, extending the computational capabilities of chemical systems.

Findings

Differentiable optimization can discover sparse, functional CRNs.

The method can implement oscillators and chemical computing devices.

Regularization helps in finding meaningful network structures.

Abstract

We present a differentiable formulation of abstract chemical reaction networks (CRNs) that can be trained to solve a variety of computational tasks. Chemical reaction networks are one of the most fundamental computational substrates used by nature. We study well-mixed single-chamber systems, as well as systems with multiple chambers separated by membranes, under mass-action kinetics. We demonstrate that differentiable optimisation, combined with proper regularisation, can discover non-trivial sparse reaction networks that can implement various sorts of oscillators and other chemical computing devices.